The next version…

These are some of the desired feature additions to the next version of EvalC3:

  1. Attribute selection: This may be in two forms
    1. Optimising the set of attributes in use by using Solver’s genetic algorithm, to optimise any one of these performance measures:
      1. Diversity: The proportion of all possible configurations that are represented within the data set
      2. Consistency: The consistency of each configuration’s relationships to the outcomes in the data set
      3. The best possible combination of Consistency and Diversity
    2. Generating a tabular display in the same form as a Truth Table used in QCA, showing cases sorted by similarity of their configurations of attributes
    3. Look here for more on Feature selection
  2. Additional measures of case similarity and difference. After selecting a specific case of interest users will be able to find other cases that fit whichever one of these categories is of the most interest
    1. Case(s) with most similar attributes and same outcome
    2. Case(s) with most similar attributes but different outcome
    3. Case(s) with most different attributes but same outcome
    4. Case(s) with most different attributes and different outcome
  3. A confidence measure for models produced from data sets that have missing values
    1. At present the search process (manual or algorithmic) treats all missing values as non-presence of the attribute being tested. This seems to lead to a prediction model that is maximally conservative, possibly understating its performance if missing values were found.
    2. Another possibility, not yet implemented, is to find the most optimistic prediction model, by assuming that all missing values for a given attribute of interest are those that are part of the model being tested
    3. Both of these assumptions might be testable by  a third strategy, which inserts  randomly chosen values  in place of missing values. Do any of these better than the optimistic model or worse than the conservative model?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s